Road Map

In the summer of 2009, the U.S. Ninth Circuit Court of Appeals overturned the U.S. Tax Court decision in Xilinx v.Commissioner (hereinafter “Xilinx”).[1] In essence, the Ninth Circuit decided that the rule in the U.S. Internal Revenue Code § 482 regulations that requires all costs to be shared in a related party cost-sharing agreement, an agreement which allows related parties in a joint venture to fix percentages of cost sharing between the two entities effectively allowing the organization to control tax allocations among lower-tax rate jurisdictions, trumps the arm’s-length standard provided in another provision of that section based on the legal maxim that the specific overrides the general.[2] Xilinx disagreed with the Ninth Circuit decision and filed a motion with the court for either a rehearing of the case or a rehearing of the case en banc. In an unprecedented move, on March 22, 2010, the Ninth Circuit reversed itself by issuing new opinions affirming the earlier tax court ruling. [3] In this paper, I will discuss the reasoning and history behind § 482 and look at the court’s application of it in Xilinx. I will also analyze the impact of the court’s decision on the arm’s-length standard and foreign tax treaties and discuss alternatives to the arm’s-length standard transfer pricing method.

What Is § 482 and Why Does It Matter?

Transfer Pricing

Transfer pricing refers to the pricing of goods or services transferred within a multinational organization. This occurs when one company (i.e., a parent company) sells a product to another wholly owned company (i.e., a foreign subsidiary). Because the prices are set internally within the organization, the typical market mechanism that establishes prices for such transactions does not apply. The company is able to decide how it would like to allocate profits among entities. The main issue in this scenario, then, is whether the tax authorities of each jurisdiction are able to tax income equivalent to that which they would have been able to tax had the transaction occurred under an ordinary arm’s-length scenario between two unrelated companies.

Transfer pricing is utilized as a tax avoidance strategy wherein “income that is earned by a high tax rate entity [is] somehow realized instead by a closely related entity that pays tax at a lower tax rate. In this way, the total tax burden faced by a group with common interests can be reduced, sometimes dramatically.”[4] Transfer pricing rules of § 482 are “an attempt to prevent earnings shifting schemes from significantly reducing the total tax obligations faced by corporate groups operating internationally.”[5]

For example, suppose Company X, a U.S. corporation, makes computer chips. Most of its sales are realized abroad through a wholly owned subsidiary, Company Y. Each chip costs $50 to manufacture and is sold to its clients abroad for $90 by Company Y. If Company Y is located in a low-tax foreign jurisdiction, then Company X is able to choose in which jurisdiction its taxable income is realized simply by controlling the sales price. Company X can realize the bulk of its income in the foreign jurisdiction by charging company Y only $55 per chip, making a token profit of only $5 per chip in the U.S. Company Y will realize a profit of $35 per chip ($90 less the $55 it paid to Company X for each chip). As long as the IRS does not challenge this pricing system, for the purpose of allocating income from the sales of each chip for determining income tax owed, the profits will be shared by Companies X and Y, with only $5 of profit realized in the U.S. and $35 in Company Y’s low-tax jurisdiction for each microchip manufactured and sold. 

Therefore, by using the concept of transfer pricing, Company X has incentive to realize as much income as possible in the lower tax rate jurisdiction. Eventually Country X will get to tax when the profits from Company Y are repatriated in the form of dividends, but this still gives Company X a method to defer taxes and compound its income at a higher effective after-tax rate of return.[6]

This type of arrangement is supplemented with a cost-sharing agreement (CSA). Controlled taxpayers can avoid the rules on intangible ownership by making a CSA.[7] This allows participants to share the costs incurred by all of them in one or more research and development projects, and they are in turn treated as jointly owning any intangibles that result from the projects. The participants make payments to each other throughout the course of the project to adjust their costs for the project to pre-agreed to proportions as set forth in the CSA. In this example, Company X would form a CSA with Company Y to determine its individual percentage shares of costs. Therefore, not only can Company X specifically allocate its profits to control how and when it pays its taxes to each jurisdiction, but it can also allocate its costs to control the amount of tax deductions received by each entity. In this clever way, Company X can potentially take higher deductions now and incur taxes for the corresponding profits later.

Transfer pricing becomes an issue because of the existence and tax treatment of different corporate entities that are related through ownership. In the example above, U.S. tax law does not consolidate the accounts of companies X and Y, but rather treats these companies as two different entities for tax purposes, and the total profits accruing to the corporate family (i.e., $40 per transaction) are not considered significant for U.S. tax purposes. But if both companies are owned by the same people, they will care only about the total return after taxes from the sale of microchips. Therefore, the economic reality is that Company X will make a total profit of $40 per transaction regardless of how its income is formally allocated. It makes sense, then, for Company X to realize the most amount of profit in the lowest-tax rate jurisdictions in which it operates. “Wherever there are different rates of taxation in different jurisdictions, multinational companies will have an incentive to engage in strategic transfer pricing by shifting profits among multiple entities to the lower tax jurisdiction without lowering the overall economic profit.”[8]

The Arm’s-Length Standard

Multinational enterprises (MNEs) have compelling business reasons that motivate the form of the transaction underlying the transfer pricing problem. In addition to wanting to take advantage of economies of scale and the reduced transaction costs that dealing with related corporate entities affords, “dealing with related entities also allows companies to keep trade secrets and other corporate knowledge within the corporate family. MNEs may legitimately want to limit the number of people who have access to the fruits of their research and development by keeping their intellectual property from the open market.”[9] As Charles McLure writes, “An influential body of literature emphasizes that modern corporations exist precisely because of the difficulties of relying on market transactions under certain circumstances. When know-how or other intangible assets are important inputs, there are ‘transactions difficulties.’”[10]

Intangible assets create a particularly difficult problem for tax administrators. In the early 1960s, the IRS assumed the practice of using prices from like transactions between unrelated parties to allocate income from transactions among related entities wherever possible. However, intangible property is often unique, and there are often not any comparable transactions among unrelated entities for the IRS to use as a basis for establishing the prices (and thereby determining the appropriate income allocations) from the contested related-party transaction.[11] 

As Professor Richard Caves emphasizes:

“The transactional model of the MNE holds that international firms arise in order to evade the failure of certain arm’s-length markets, especially those for intangible assets. Premier among those assets is the knowledge that represents new products, processes, proprietary technology, and the like. Thus, theory implies and empirical evidence confirms that MNEs appear prominently in industries marked by high expenditures on research and development and rapid rates of new product introduction and productivity advance….  MNEs arise because of shortcomings in arm’s-length markets for intangible assets, and the statistical evidence establishes the prominence of MNEs in high R & D industries.”[12]

The History of Transfer Pricing and the Development of § 482

The Commissioner of the Internal Revenue Service first received authorization to allocate income and deductions among related corporate entities in the War Revenue Act of 1917: “Whenever necessary to more equitably determine the invested capital or taxable income.”[13] Then, The Revenue Act of 1921 went even further and allowed the Commissioner to prepare consolidated accounts to correct income received from commonly controlled trades or businesses: “for the purpose of making an accurate distribution or apportionment of gains, profits, income, deductions, or capital....”[14] Legislative history suggests that this provision was enacted to provide the Commissioner with an effective means to contest tax avoidance and prevent companies from shifting income to tax-favored foreign subsidiaries through manipulative transfer pricing practices.[15] For example, the House Ways and Means Committee’s Report states that “subsidiary corporations, particularly foreign subsidiaries, are sometimes employed to ‘milk’ the parent corporation, or otherwise improperly manipulate the financial accounts of the parent company.”[16]  

The Revenue Act of 1928 turned its back on the consolidation method of the 1921 Act, substituting for the phrase “consolidate the accounts” a rule authorizing the Commissioner to “distribute, apportion or allocate gross income or deductions between or among such trades or businesses, [as] necessary in order to prevent evasion of taxes or clearly to reflect the income of any such trades or businesses.”[17]

In 1935 the IRS issued the first transfer pricing regulations under the predecessor to § 482. These regulations unambiguously established the arm’s-length standard, the norm used to this day by the Commissioner to analyze and allocate the income from transfer pricing transactions. The provision gave the Commissioner authority to make adjustments “expressly predicated upon his duty to prevent tax avoidance and to ensure the clear reflection of the income of the related parties.”[18] 

By the 1960s, Congress saw that its efforts to control transfer pricing were not working properly. U.S. companies were succeeding in shifting U.S. taxable income to their foreign subsidiaries, fueled in their efforts by court decisions favoring MNEs in the intervening decades.[19] For example, in Seminole Flavor Co. v Commissioner, instead of looking at whether the distribution agreement would have been concluded at the same price had it been negotiated at arm’s length, the court focused on whether payments made under a distribution agreement were “fair or reasonable.”[20] Other courts at this time also adopted similarly worded fairness tests and did not try to find comparable transactions in order to arrive at an arm’s- length standard price for the contested transactions.[21] 

The Treasury complained in 1961 that “transactions between related domestic and foreign entities cause extreme administrative difficulties.”[22] Congress reacted by instructing the Treasury Department to issue new regulations under the broad authority granted to it under ­­­     § 482.[23] According to Reuven Avi-Yonah, these regulations, which remained in effect until 1993, attempted to clarify the application of the arm’s-length standard (ALS):

“[T]he regulations attempted to establish rules for applying the ALS to specific types of transactions, but with different degrees of specificity. For services, the regulations merely recited the ALS without any guidance as to its application in the absence of comparables. For intangibles, the regulations contemplated a failure to find comparables. They list twelve factors to be taken into account, but without establishing any priority or relative weight among them.”[24] 

Of the three methods set out for analyzing transactions, the comparable uncontrolled price method (i.e., the price at which unrelated parties have conducted similar transactions) is described by the Treasury Department as “likely to achieve the highest degree of comparability of any method potentially applicable to a transfer of tangible property.”[25]   

But comparable transactions were often “either absent or misused when transfers of intangible property [were] at issue.”[26] Intangible property, such as a license for a patent or some other sort of intellectual property, is often unique, and this uniqueness is often where it derives a substantial part of its value. As such, when intangibles are involved it is less likely that there will be transactions involving comparable properties for the Commissioner to utilize to analyze transfer pricing transactions. As a result, “analyzing transfer-pricing transactions involving intangible property has been and continues to be difficult and controversial.”[27] 

Congress amended § 482 in 1986 in an effort to remedy transfer pricing problems associated with licenses of intangibles and invited the Treasury Department to create new transfer pricing regulations. In response, the Treasury issued regulations in 1992, 1993 and 1994, all of which had substantial differences.[28] “The most noteworthy feature of the 1993 regulations in comparison to earlier versions of regulations under § 482 was the emphasis on comparability, and the resulting flexibility. The final regulations adhere to this emphasis and in some cases increase it.”[29] The best-method rule is one of the sources of this increased flexibility, which gives the Commissioner discretion in determining which method to use: “The arm’s-length result of a controlled contraction must be determined under the method that, under the facts and circumstances, provides the most reliable measure of an arm’s-length result. Thus, there is no strict priority of methods, and no method will invariably be considered to be more reliable than others.”[30] 

It is clear that “the changes made might have been directly influenced by, among others, U.S. trading partners, in particular the members of the Organization for Economic Cooperation and Development (OECD).”[31] Indeed, by reaffirming the arm’s-length standard in 1995 the OECD signaled an ongoing international endorsement of that standard, and its endorsement is often referred to as evidence of an international consensus in its favor.[32] Critics of proposals to replace the arm’s-length standard with some other transfer pricing methodology maintain that the OECD’s endorsement makes clear that the current international consensus requires the U.S. to continue its use of standard the arm’s-length approach. Changing from the arm’s-length standard in spite of this international consensus might burden MNEs with different inquiries from the tax authorities in the U.S. and elsewhere. They would then need to maintain at least two sets of books: one following the accounting requirements of the arm’s-length standard and another following the accounting requirements of an alternative standard.[33] 

The Arm’s-Length Standard Today

At present the regulations under § 482 state that, “In determining the true taxable income of a controlled taxpayer, the standard to be applied in every case is that of a taxpayer dealing at arm’s length with an uncontrolled taxpayer.”[34] Furthermore, “a controlled transaction meets the arm’s-length standard if the results of the transaction are consistent with the results that would have been realized if uncontrolled taxpayers had engaged in the same transactions under the same circumstances (arm’s-length result).”[35] The arm’s-length standard necessitates that each party calculate its profits separately; that is, without consolidating its accounts with related taxpayers and demands that related party transactions be priced as if unrelated parties had entered into them. “The current regulations adopt a market-based approach, distributing income among related parties the way a free market would distribute it among unrelated parties.”[36]

Criticisms to the Arm’s-Length Standard and Alternative Methods

There are, however, critics of the arm’s-length standard. These critics claim the traditional arm’s-length standard has failed, because it does not reflect economic reality.[37]  The subsidiaries are fully under the control of the multinational which logically considers the tax implications in setting prices; naturally, MNEs exist because of market and non market advantages that are derived from their structure. It avoids (internalizes) transaction costs, which increases efficiency in raising capital, advertising products, achieving economies of scale, and protecting valuable intangibles. Thus, if one applies a market rate of return separately to each of the components of the multinational, the result is less than the actual return of the organization as a whole. This residual cannot be assigned to any component. Any transfer pricing rule which arbitrarily assigns the residual to one part of the organization distorts economic reality.[38]

Practically, the arm’s-length standard creates a climate of uncertainty in the absence of clearer rules. Neither the taxpayer nor the IRS can know in advance the likely revenue outcome in a transfer pricing case. The inability to forecast the taxes on international ventures with reasonable certainty may discourage taxpayers from undertaking such ventures despite higher potential returns.[39] 

Other Allocation Methods

Critics of the arm’s-length standard have suggested alternative methods to transfer pricing. Among the most popular is the “unitary” standard, which looks to the total profits earned by firms that are under common control and engaged in similar or related activities.[40] Under this approach, related entities would be treated as parts of the same company – as composing a “unitary” business. The income that all transactions between these parties generate would be allocated using a predetermined formula consisting of three elements: sales activity in each jurisdiction, property in each jurisdiction, and payroll in each jurisdiction. For each variable, the ratio of the MNE’s income earned in the taxing state to the income earned worldwide is calculated, i.e., sales in state divided by sales worldwide, property in state divided by property worldwide, and payroll in state divided by payroll worldwide. 

Developing these terms into a formula, the percentage of income allocated to a particular state is equal to:

[ [State Sales]/[All Sales] + [State Property]/[All Property] + [State Payroll]/[All Payroll] ] / 3 

= % of Income Allocated to State

In many U.S. states where this formula is used it is common to weight sales more heavily than the other two factors. In Canada, most of the provinces use a formula with only two equally weighted apportionment factors – payroll and sales.[41] 

Commentary on the Unitary Method

The unitary method is a more accurate depiction of profit sharing because it considers the theory that activities in each of the jurisdictions where an enterprise operates contribute to its overall profit.[42] The system apportions that profit to the jurisdictions based not on a manipulable internal pricing scheme, but on a mathematical “rough approximation” of the income “that is reasonably related to the activities conducted within the taxing jurisdictions.”[43]

The unitary system is not without its problems. The success of this system requires uniformity among jurisdictions. A multinational company may face the unitary method in four of its six jurisdictions while the other two jurisdictions seek to tax a local part of the business as a separate operation. The result can be the taxation of the same income by two or more states.[44] In theory the unitary method appears to be a successful method of income allocation, while in practice it may be difficult to implement on a world-wide basis.

Xilinx Facts

Xilinx Inc. (Xilinx) is in the business of researching, developing, manufacturing, marketing, and selling field programmable logic devices, integrated chip devices, and other development software systems.[45] During the years in issue, Xilinx was a parent of a group of affiliated subsidiaries including Xilinx Holding One Ltd., Xilinx Holding Two Ltd., and Xilinx Ireland (XI). XI was established in 1994 as an unlimited liability company under the laws of Ireland and was owned by Xilinx Holding One Ltd., and Xilinx Holding Two Ltd. (i.e., Irish subsidiaries of Xilinx). XI was created to manufacture field programmable logic devices and to increase Xilinx’s European market share. It manufactured, marketed and sold field programmable logic devices, primarily to customers in Europe, and conducted research and development. 

In  1995, Xilinx and XI entered into a cost-sharing agreement (CSA). The U.S. parent had more than $100 million in employee stock option (ESO) benefits afforded to all of its U.S. employees.  These ESOs generated compensation deductions for Xilinx under I.R.C. § 162. No ESO amount was reflected as a shared cost in the CSA. The Irish subsidiary compensated the U.S. parent for ESO costs related to employees of the Irish subsidiary.[46]

The Commissioner issued notices of deficiency for the tax years involved claiming that Xilinx should have included in its CSA with XI the costs of U.S. ESOs, thus reducing Xilinx’s § 162 deduction and thereby increasing its taxable U.S. income. The Commissioner’s determination meant that Xilinx owed substantial tax deficiencies and accuracy related penalties under 26 U.S.C. § 6662(a).[47]

The Courts Application of the § 482 Regulations and the Arm’s-Length Standard and Treaties

At trial, the U.S. Tax Court found for Xilinx. It held that the arm’s-length standard is the applicable method to determine the allocation of costs in a CSA between related parties. It determined that Xilinx’s allocation of ESOs met the arm’s-length standard mandated by § 482. In other words, the court decided that unrelated parties in a similar transaction would have followed the same allocation of costs upon which Xilinx and XI agreed.[48]

The Commissioner appealed and brought the case to the U.S. Ninth Circuit Court of Appeals in Xilinx Inc. v. C.I.R.[49] The Ninth Circuit court reversed the tax court and determined instead that “all costs,” as proscribed by Section 1.482-7(d)(1), are to be shared in a CSA among related parties. Therefore, they concluded that Xilinx should have included as a cost in the CSA with its Ireland subsidiary the ESO costs of the U.S. parent related to its own employees. The court reasoned that Section 1.482-7(d)(1) effectively overruled Section 1.482-1(b)(1), which requires related parties to follow the arm’s-length standard. It used the legal maxim that the more specific statute overrules the more general. As applied to this case, a company that is engaged in a related party transfer pricing transaction should share “all costs” (including ESOs) rather than follow the arm’s-length standard (which the court agreed would have supported Xilinx’s decision to not share the ESOs).[50] This meant that Xilinx would have to adjust its relevant U.S. tax filings and pay penalties.

The second issue the court analyzed dealt with the impact this new interpretation of § 482 had on the Ireland-U.S. tax treaty, which states that the treaty incorporates the arm’s-length standard from U.S. tax law.[51] The court did not find that trumping the arm’s-length standard violated the Ireland-U.S. tax treaty because of the presence of article 1(4) thereof (the so-called savings clause) ‘‘because the treaty expressly allows a contracting state to apply its domestic laws to its own citizens, even if those laws conflict with the treaty.”[52]

Soon after this decision, Xilinx moved for a rehearing en banc. In a move that demonstrated the importance of this issue in the tax community, the Ninth Circuit withdrew its opinion and reissued a new opinion on March 22, 2010.[53] The court, in a two to one opinion, held that Xilinx correctly used the arm’s-length standard and owed no deficiencies to the Commissioner.[54]

The court reasoned that it was to base its decision between two alternatives: that the specific controls the general, or that it should “resolve the ambiguity based on the dominant purpose of the regulations.”[55] The court stated that to follow the general rule that the specific controls the general would be akin to “converting” a canon of construction into something like a statute. Instead it decided that “purpose was paramount” and that the purpose of the regulations is “parity between taxpayers in uncontrolled transactions and taxpayers in controlled transactions.” If 7(d)(1) were to trump the arm’s-length standard then that purpose would be frustrated and would not allow Xilinx to deduct all its stock option costs, thereby taking away its tax parity with an independent taxpayer.[56]

The court further affirmed in dicta the importance of the Ireland-U.S. tax treaty and its acceptance of the arm’s-length standard as the standard international method for transfer pricing in a related party cost-sharing agreement. The court did, however, specify that it did not intend in this case to decide whether treaty obligations “constitute binding federal law enforceable in United States courts.”[57]

Commentary on Xilinx

An interesting effect of this decision is its impact on U.S. tax treaties. The District Court justified its first decision by hanging it on the presence of article 1(4) of the Ireland-U.S. tax treaty.[58] This so-called savings clause allows a contracting state to apply its domestic laws to its own citizens, even if those laws conflict with the treaty.[59] However, opponents argued that the court was misreading the savings clause and that it did not apply to foreign joint venture cost-sharing agreements.[60]  They claim that article 1(5)(a) of the treaty is a carveout of the savings clause in article 1(4), which lists article 9(2) as not being subject to the savings clause.[61]

In the end, the Ninth Circuit reaffirmed the tax court but did not opine concerning the interpretation of the savings clause.[62] Does this leave open the validity of international tax treaties between the U.S. and foreign jurisdictions? As it stands, Xilinx means that the Ireland-U.S tax treaty has effect and its support of the arm’s-length standard is upheld. However, it seems that the validity of a tax treaty now may only be as strong as the degree to which the Commissioner likes the overall effect the treaty has in filling the government’s coffers. The Commissioner has yet to file for certiorari regarding Xilinx to the U.S. Supreme Court and there is no indication that he is headed in that direction.

As for the arm’s-length standard, barring the grant of a petition for certiorari, it is back and it is king. With the onset of globalization, companies went multinational and it became cost-effective to enter into cost-sharing agreements with foreign subsidiaries to advance the development of products, especially those products that were intangible in nature. For companies like Xilinx, millions of dollars can be saved through the use of a related party CSA. These cost savings stimulate research and development and international joint ventures. The arm’s-length standard supports the global business environment and encourages the research and development that is necessary to further innovation and create new and better products and technologies for our global society. 

Author’s Comments

The Ninth Circuit court correctly chose the arm’s-length standard (ALS) over the “all costs” method of allocating costs among a related party joint venture. This decision affirms the primacy of the ALS as a “foundation of the U.S. transfer pricing regime and rejects the IRS’s attempt to interpret its own rules in conflict with the [ALS].”[63] The decision will surely impact virtually all transfer pricing issues, regardless of whether those issues involve cost sharing.[64] The court specifically found that there was a conflict in the regulations and that the ALS controlled. “Thus, much to the IRS’s chagrin, evidence of what unrelated parties actually do will remain a staple of any § 482 analysis.”[65]

Incidentally, this decision also curtails the IRS’s ability to promulgate rules in the transfer pricing arena mandating a certain result without due consideration for what unrelated parties actually do. “Arm’s length results cannot be achieved by fiat.”[66] This arms taxpayers with a strong arrow in their quivers with which to respond to adjustments that are not based on actual transactions between unrelated parties.

The decision to continue to follow the ALS was correct because it saves billions of dollars in unforeseen costs for companies with cost-sharing agreements (CSA) and possibly trillions of dollars in non-CSA intercompany transactions.[67] For example, after the Ninth Circuit’s original decision reversing the U.S. tax court, Cisco Systems, Inc., one of the 33 amici representing interested U.S. corporations, reported additional liabilities due to the Xilinx decision in excess of $720 million.[68] Furthermore, the multi-billion dollar savings provided for by this decision under the cost-sharing regulation at issue fails to include the impact that this decision will have on intercompany transactions under other regulations. For example, the “all costs” method “… impugns the accepted interpretation of other regulations under § 482 that are potentially applicable to trillions of dollars of intercompany transactions.”[69]

This decision is also correct because it supports the U.S. treaty network and global business.[70] The ALS provides the international benchmark for transfer pricing (i.e., allocating tax consequences for transactions between related parties): “a standard designed to prevent variations in tax treatment from impeding international trade and economic development.”[71] The standard allows companies under common control, yet subject to different taxing authorities, to adjust intercompany income and deductions associated with transactions to reflect the allocations that would have resulted had independent companies engaged in the same transaction.[72] The “all cost” method would bring U.S. tax law into conflict with treaty provisions that are premised on the reciprocal application of this standard. Furthermore, relying on the ALS, thousands of U.S. companies and their overseas affiliates engage in transactions valued in the trillions of dollars. A change from the ALS would disrupt the settled expectations of those companies, thwarting the goal of consistent transfer pricing resulting in international double taxation. Instead, this decision appropriately reinforces the fundamental premise of the international transfer pricing regime and sustains trust in the U.S. as a treaty partner – the partner who first convinced the world to adopt the arm’s length analysis.[73]